ottersnap-mcp
以下摘录自该仓库的 README(片段) · 在 GitHub 读完整 README →
MCP (Model Context Protocol) server for **Glint Render** — the web rendering & evidence platform by Moyu. — give Claude, Cursor, Codex and other AI agents **eyes on any web page**: pixel-perfect screenshots, print-ready PDFs and branded OG images from natural language.
Quick start
Get a free API key at ottersnap.com/dashboard (100 renders/month, no credit card), then add to your MCP client config:
Or install globally:
Tools
| Tool | What it does |
|---|---|
| render_screenshot | Renders any URL and **returns the image visually** — the agent can actually read the page. Supports full-page, retina 2×, dark mode, element hiding, cookie-banner blocking, iPhone/iPad/Android presets. |
| render_pdf | URL or raw HTML → print-ready PDF (A4/Letter/Legal, landscape, margins, header/footer with page numbers). |
| create_og_image | Branded 1200×630 social card, returned visually. Seven themes. |
| extract_page | URL → structured JSON (title, description, headings, links, images, word count) or clean Markdown — built for RAG and agent pipelines. |
| ai_extract | URL + natural-language prompt → structured JSON extracted by LLM. |
| code_image | Code snippet → macOS-window syntax-highlighted PNG. |
| check_usage | Remaining renders for the configured key. |
| workflow_create | Create a **workflow** — an automation that runs steps (screenshot, extract, fetch any public API, AI, email, webhook) on a schedule, when a monitored page changes, or via webhook. |
| workflow_list | List your workflows: trigger, steps, enabled state, last run. |
| workflow_run | Run a workflow immediately and report every step's result. |
| workflow_runs | Recent run history of a workflow. |
Example prompts
- "Screenshot github.com/trending and tell me which repos mention Rust"
- "Create a workflow that watches my landing page and emails me an AI summary whenever it changes"
- "Take a dark-mode full-page capture of my landing page and check the footer renders"
- "Extract every heading and link from https://example.com/docs as markdown"
- "Scrape the product name, price and rating from this URL into JSON"